code and test set for article "<b>Single-shot full-Stokes imaging through scattering media</b>"

This repository provides the source code, test datasets, trained model weights, and predicted results associated with Figures 4 and 5 in the main manuscript. These figures illustrate AI-assisted metasurface imaging through scattering media. Due to differences in image naming conventions between the two experiments, the <code>split_data.py</code> script is slightly modified for each figure. We recommend running the code on a cloud-based environment to avoid potential dependency and configuration issues.本仓库提供了与主稿件中图4和图5相关的源代码、测试数据集、训练好的模型权重以及预测结果。上述两幅图展示了基于散射介质的AI辅助超表面成像技术。由于两项实验的图像命名规范存在差异，针对每个图都对<code>split_data.py</code>脚本进行了小幅修改。我们建议在云端环境中运行代码，以规避潜在的依赖项与配置问题。